The following list contains open projects I am offering to supervise in the form of Master's theses. Reduced alterations of given projects might also be tackled for Bachelor's or other forms of research projects. If you are interested in a given topic, please contact me and we can schedule a meeting in person or via Skype. Also, if you have own project ideas, write a brief proposal and we discuss them.
Predicting memes on Reddit
On Reddit, a popular social news aggregation website, memes are omnipresent. In this project, we are interested in predicting popular meme responses to submissions and comments. For example, if someone posts a funny comment, the goal would be to automatically reply with a witty GIF meme that also gets perceived well by the community. We could approach this task by utilizing machine learning and implementing a Reddit bot that automatically replies GIFs based on our predictions. Reddit provides a natural evaluation platform as we can directly judge the plausibility of our algorithm by looking at the up- and downvotes of our automatically produced posts.
Requirements: Python, numpy, scipy, scikit-learn, machine learning basics
Sequential data analysis framework
Sequential data emerges in a multitude of scenarios; for example, if someone navigates between websites on the Web. There exists a variety of methodological approaches to study this kind of data such as sequential pattern mining or Markov chain modeling. However, there is a lack of a unified framework that allows researchers to study sequential data with this variety of methods. In this project, your goal would be to initiate a Python framework covering most of the basic sequential data algorithms coupled with an understanding of given methods. You would then also apply these methods to sequential data and analyze the observed results.
Requirements: Python, numpy, scipy, basic understanding of sequential data methods
Markov chain models
The basic Markov chain model that stochastically models transitions between states in sequences. In literature, several adaptions of this simple model have been proposed such as higher-order Markov chain models, varying-order Markov chain models or semi Markov chain models. The goal of this Master's thesis would be to review the most popular alterations and summarize them in the thesis. Then, given some empirical sequential dataset such as human Web navigation, the different methods should be applied to the data and results should be compared and analyzed.
Requirements: Basic understanding of Markov chain models, knowledge and interest in statistics